A Fast Algorithm for Scheduling Detection-and-Rescue Operations Based on Data from Wireless Sensor Networks

The need for the search, detection and rescue of disaster survivors arises in many emergency situations in military and civil applications. Suppose a number of people are trapped in ruins after an earthquake or tsunami. Their medical condition depends on their location, detection time and the time of the rescue operation. In order to efficiently detect and perform the needed rescue operations, a network of wireless sensors is used which provide acoustic, seismic, electromagnetic, gravimetric and other information. The information is processed automatically to yield prior probabilities of location and expected rescue times for each potential target. The acquired information from the sensors is imperfect because under extraordinary and severe circumstances, two types of errors may occur: (i) a "false-negative detection test” – it is a case when a target is overlooked during the test; and (ii) a "false-positive detection", or "false alarm" – when a not-a-target location is wrongly classified as a sought target. Therefore, non-zero probabilities of overlooking a hidden target and a "false alarm" exist. We suggest a two-phase solution to the problem of scheduling detection and rescue operations. First, the disaster area is divided into sub-areas and available rescue teams and sensors are assigned. Second, a schedule is found for the rescue teams to perform the rescue operations (in parallel). We seek to find the best coverage of the disaster sub-areas served by rescue teams and to schedule the search-and-rescue operations in each sub-area while minimizing the search-andrescue time and maximizing the number of saved lives within a given search time limit. The problem is formulated as a non-standard two-stage assignment / scheduling problem and a fast combinatorial real-time algorithm is suggested.